Professional Documents
Culture Documents
Abstract
The mechanical problem discussed in this paper focuses on the stress state estimation
in a composite laminate in the vicinity of a free edge or microcracks. To calculate these
stresses, we use two models called Multiparticle Models of Multilayered Materials (M4). The
first one can be considered as a stacking sequence of Reissner-Mindlin plates (5 kinematic
fields per layer), while the second is a membranar superposition (2 fields per layer plus a
global one). These simplified models are able to provide finite values of interfacial stresses,
even on the free edges of a structure. The current paper consists of validating the M4 by a
finite element analysis through describing the stress fields in both a (0,90)s laminate in
tension (free-edge problem) and a transversally microcracked (0,90)s laminate. A comparison
of the various energy contributions helps yield a mechanical perspective: it appears possible
to define an interply energy as well as a layer energy, these energies expressing the FE 3D
reality.
Keywords:
composite laminate - modelling - multiparticle - finite elements free edge - 3D
stresses - energy
1. Introduction
Due to the large difference in anisotropy of two consecutive plies, high interlaminar
stresses (at the interplies and especially in the vicinity of a free edge) are induced in cross-ply
laminates and lead to damage such as delamination. Classical lamination theory is not able to
calculate these out-of-plane stresses. Many studies have sought to overcome this lack of
classical lamination theory by calculating the interlaminar stresses in a laminate subjected to a
tensile loading. A bibliography reviewing each of these studies in detail is detailed in this
1
paper. They include numerical and analytical studies, the finite difference technique, the finite
element method, boundary layer theories with corrective terms and models with a kinematic
field per layer, that we named multiparticle models.
The models proposed herein, called Multiparticle Models of Multilayered Materials (M4),
clearly belong to this last family of models. Their construction will be summarily described.
In order to validate our models, two examples corresponding to different boundary conditions
are treated. The first one consists of analysing the free edge problem in a (0,90)s laminate
submitted to tension. The second considers, in the same laminate, microcracks present within
the 90 plies. These two examples, as well as all materials constants and calculation
hypotheses, are depicted in the following section.. The use of such cross-ply allows us to
avoid 3D calculations by admitting plane deformation assumptions, and hence 2D
calculations. The study of a (,-)s will enable drawing the exact same kind of observations
and conclusions, but with 3D meshing (see Figure 12, for example).
Once the validity of the finite element calculations (numerical convergence and study of
singularities) has been ensured, a comparison of the two approaches is drawn in the papers
final section, focusing on:: 3D stress fields, interface forces, and proposing energy-related
considerations. In the case of free edge problem, our results also make reference to some of
Pagano's works (1978).
In this section, we will present some of the methods available for calculating
interlaminar stresses in any laminate with straight free edges. These consist of: finite element
procedures, boundary layer theories and multiparticle models.
Wang & Crossman (1977) used a finite element numerical procedure based on a
displacement formulation. The field singularities between two plies and near the free edge are
highlighted. For the crossplies ((0/90)s, (90/0)s), they established a description (on the mid-
plane) of zz that displays a different sign between the two cases. These various curves for
zz may serve to justify the distinction in the two laminates' behaviour regarding damage at
this interface. At the 0/90 interface, shear stress conditions are also different: for the (0/90)s
laminate, a singularity seems to exist at the free edge, which is not the case for the (90/0)s
laminate. Raju & Crews (1981) investigated the (,-90)s family. With a refined polar mesh,
they were able to determine the stress singularity order. This singularity was studied by a
2
number of authors (Wang & Choi (1982a) , Leguillon (1998)), etc.) and has often been
identified using a logarithmic expression. Shah & Murty (1991) modelled the laminate as a
combination of three distinct regions: quasi-3D elements close to the free edge, linear
elements over the entire plate, and transitional elements between these two regions. Robbins
& Reddy (1993) developed a 2D layerwise, displacement-based finite element model of
laminated composites that assumes a per-layer distribution of the displacement field (1D
elements on the thickness).
Because classical lamination models yield an accurate approximation of fields except
in the vicinity of edges, it appears altogether natural to superimpose corrective fields whose
values are only significant at the edges (boundary layer theory) onto these lamination fields.
Let us first point out the asymptotic development technique by Lcuyer (1991) and the
Fourier series development of Allix (1992). Wang & Choi (1982b) presented a formulation of
stress functions based on complex variables. The boundary layer asymptotic stresses are
characterised by introducing a stress intensity factor dependent upon geometrical variables
(e.g. laminate thickness, number of plies), stacking parameters (fibre orientation, stacking
sequence) and environmental conditions (temperature and relative humidity).
Our attention is now turned to defining a family of models we call multiparticle models; these
consider the existence of many material particles at a single geometrical point, i.e. one particle
(or one kinematic) per layer (whereas classical lamination theory involves only one kinematic
over the entire thickness). Nevertheless, they are all 2D models and, as such, can be viewed as
plates or membranar superpositions linked together by interface forces. Garett & Bailey
(1977) developed a model that enables solving transverse-cracking problems in a (0/90)s
laminate. This model, called shear-lag, has been widely used by other authors (Caron et
Ehrlacher (1997,1998), Carreira(1998), Steif (1983). The most complete multiparticle model
is the local model of Pagano (1978). He proposed a laminated composite theory based on the
Hellinger-Reissner variational principle (1950): the membranar stresses in each ply are
written as first-order polynomials, and the shear and normal stresses are then obtained by
integrating the 3D equilibrium equations. The various studies of Caron & Ehrlacher (1997),
Chabot (1997), Naciri et al. (1998), 1998), Caron et al. (1999), Carreira (1998), (referring to
the M4) take their inspiration from Paganos model (1978). However they are aimed at
proposing more simplified approaches, which serve to derive analytical solutions.
3. Presentation of the two problems : free edge and micro-cracking boundary conditions
3
Each ply is considered as macroscopically homogeneous and monoclinic, and represented by its elastic
constants : El , Et , En longitudinal, transverse and normal Youngs moduli, Gij shear moduli
The first example consists of the free edge problem in a finite width (0,90)s laminate
under uniaxial tension (see figure 1), where e denotes the ply thickness, and 2b the width.
Using the relation b=8e as a base asumption, a uniform displacement is then imposed on
x=a edges. For the stacking sequence studied herein, the problem is independent of x and
due to symmetry, the problem can be reduced to an analysis of the shaded cell in the (y,z)
plane.
fig 1 : The free edge problem in a (0,90)s laminate under uniaxial tension
The second example studied, which depicts more severe stress gradients conditions, is
transverse cracking in a (0,90)s laminate under uniaxial tension in the x-direction (see Figure
2). A periodic cell (in the (x,z) plane) is highly representative of this problem and a mean
distance (2h) between two consecutive cracks can in reality be experimentally observed. In
comparison with the free edge problem, only boundary conditions have changed: instead of
an out-of-plane loading, this second problem now consists of a prescribed displacement of the
0 ply, with the 90 ply remaining free of stresses. Note that we have selected b = h = 8e in
order to ensure cells of the same dimensions.
4
fig 2 : The problem of a (0,90)s laminate under uniaxial tension with 90 ply micro-cracks
The material properties and sample geometry are summarised in Table 1 below:
Table 1 : material properties (Wang & Crossman (1977)) and sample dimensions
4. The M4 construction
We introduce the following notations : x and y represent the co-ordinates in the mid-plane of the
layer, z is the thickness co-ordinate. In each layer i (i=1,n), hi , hi+ and h i are the bottom, the
top and the mid-plane z co-ordinates of the ply, respectively, and ei = hi+ hi is the thickness.
Greek alphabet subscripts correspond to {1,2} and Latin to {1,3} (except for i which
identifies the layer).
The M4 construction method (Chabot (1997)) is based upon the four steps described below.
The M4_2n+1 is the simplest one : it considers the laminate as a membranar superposition (2n
equations plus a global one, with n being the number of plies in the laminate). Resultant
forces in each layer as well as interlaminar shear stresses are taken into account, yet resultant
5
moment in the layer is not. The M4_5n can be considered as a superposition of Reissner-
Mindlin plates, in taking the shear and moment resultant, in each layer, into account, along
with interlaminar shear and normal stresses at the interfaces (5n equilibrium equations).
and the 3D equilibrium equations lead both to shear stresses 3 in the form of quadratic
N i
( x, y) = ( x, y, z) dz ; M i
( x, y) = ( z hi ) ( x, y, z ) dz
hi hi
hi+
Qi ( x, y) = 3 ( x, y, z ) dz (1)
hi
i, i +1 ( x, y) = 33 ( x, y, hi+ ) (
i 1,i ( x, y ) = 33 x, y, hi ) (3)
6
stresses are therefore independent of the z co-ordinate. The approximation order of 3 and
33 must remain consistent with the equilibrium equations (first and second order
polynomials, respectively).
The M4_5n and M4_2n+1 generalised internal forces are summarised in Figures 3.
7
(
H .R. U ,
* *
) = *
: U ( ) f .U
* * 1 * *
(
: S : d .n . U U dS T .U dS
* * d d *
)( )
2 u t
field of a symmetrical second-order tensor. is the studied object volume, its boundary,
( )
U * is the symmetrical gradient of U * , S is the compliance tensor, T d is a surfacic force
normal to .
After integration in each ply with respect to z, these associated generalised displacements are
then deduced (see Chabot (1997) for more details). They appear as weighted-average 3D
displacements. For the M4_5n, we can define the following 5n fields :
- the in-plane displacement and rotation fields of layer i, whose components are :
hi+ hi+
z hi
U ( x, y ) = i U ( x, y, z ) dz ( x, y ) = i U (x, y, z ) dz
1 12
i i
; (4)
e i2
hi e hi e
By noting W3 =
U 3 hn+ + U 3 h1 ( ) ( )
the 2n+1 generalised displacements of the M4_2n+1 are
2
identified as U i and W3 .
The M4_5n and M4_2n+1 generalised displacements are summarised in Figure 4.
The generalised strains, deduced from the generalised displacements, appear as the cofactors
i
of the generalised internal forces in the Hellinger-Reissner functional. For the M4_5n, N ,
i =
1 i
2
(
U , + U i , ) ; i =
2
(
1 i
, + i , ) ; d = i +U 3i ,
i
i ,i +1 i ,i +1 e i i e i +1 i ,i +1
D =U U
i
; D3i ,i +1 = U 3i +1 U 3i (6)
2 2
8
Fig 4a : M4_5n generalised displacements
i
For the M4_2n+1, N and i,i +1 remain associated with i and Di ,i +1 respectively,
whereby:
e i + e i +1
Di ,i +1 = U i ,i +1 U i + W3, (7)
2
9
4.3 Equilibrium equations
Reissner (1950): The elastic solution of the problem is the pair U , ( * *
)which renders the
N
i
(
, +
i ,i +1
)
i1,i = 0 N i
, + (
i ,i +1
i1,i = 0 )
( )
5n Qi , + i ,i +1 i 1,i = 0
2n+1 n e j j , j +1
(
) ( )
(8)
, + j ,1, j + n ,n +1 0,1 = 0
( )
i
M i + e i ,i +1 + i 1,i Q i = 0 j =1 2
, 2
- the behaviour of the interlaminar shear and normal stresses at interface i,i+1 :
(
Di ,i +1 = f Qi , Qi +1 , i 1,i , i +1,i + 2 , i ,i +1 , S i 33 , S i +313 ) (11)
D3i ,i +1 = f ( i 1,i
, i ,i +1 , i +1,i + 2 , S 3333
i i +1
, S 3333 ) (12)
Lastly, we can write for the M4_2n+1 :
i =
1 i
e i
S N i ; (
Di ,i +1 = f i 1,i , i +1,i + 2 , i ,i +1 , S i 3 3 , S i +31 3 ) (13)
The complete expressions for d , Di ,i +1 and D3i ,i +1 can be found in Chabot (1997).
i
5. Numerical aspects
10
This section is intended to validate our finite element results, which are our main
reference during the steps of M4 validations, and serves to introduce what we have called the
finite element generalised interface forces.
For the first mesh for example, this reflects 12 elements per ply thickness and 50 elements in
the width direction. According to figure 5, shear stress singularity at the free edge exists. In
11
effect, as the mesh becomes finer, distance to the edge decreases and shear values increase.
As a case in point, the value for a 50x12 mesh is about 35% greater than that for a 25x6
mesh. This maximum value is thereby rendered meaningless due to mesh dependence.
Because of the steepness of stress gradients at the ply interface, particularly near the free edge
(see Figure 6), it can be observed that the coarser mesh only shows convergence (i.e.
discrepancy of less than 1%) for stress values (above and below) up to y/b=0.92. In refining
the mesh (i.e. the 50x12 mesh), convergence only occurs at a distance of about 2% (between
y/b=0.98 to y/b=1) of the cell.
Two main difficulties have arisen: 1)identifying the stresses near the edge, and 2)calculating
those located exactly at the interface. In order both to overcome these difficulties and to draw
a comparison with our simplified models, we have introduced what we call the finite element
generalised interface forces, which are very similar to those introduced for the M4 (see
equations 1 through 3).
12
We are introducing here this particular force concept in order to determine the finite element
forces specially at the interfaces and to better describe these forces near the edge. Lets write
out the first two equilibrium equations of M4_5n (equation 8 left) :
i
N ( i ,i +1
, + )
i1,i = 0 ; (
Qi , + i ,i +1 i 1,i = 0 )
By summing over the first j plies, we obtain the following expressions for the interlaminar
shear and normal stresses at interface j,j+1 (no surfacic force has been prescribed) :
j j
j , j +1 = N
i
, ; j , j +1 = Qi , (14)
i =1 i =1
For a (0,90)s laminate submitted to uniaxial tension, the finite element shear and normal
stresses ( 02,90 FE , 0FE,90 and 90
FE i i
, 90 , respectively), can be deduced by deriving N 22 and Q2
numerically:
0 , 90 FE
= 23 ( x, y, z = e ) =
N 22
90
(y) ; FE
= 33 ( x, y, z = e ) =
Q20
y y
2 0 , 90
Q20 Q290
90 , 90 = 33 ( x, y , z = 2e ) = +
FE
(15)
y y
As undertaken previously for the mean shear stresses, we now conduct a convergence study
for the generalised shear stresses. The curves in Figure 7 attest to the convergence of 2FE ,
y
regardless of the mesh used, as long as < 0.98 .
b
The finite element generalised internal forces we defined are in fact more effective and
pertinent in estimating 3D stresses near the edge.
13
Fig 7 : Convergence of the finite element generalised shear stresses
be studied separately (Kassapoglou & Lagace (1986)) and the hypothesis of material being
macroscopically homogeneous is no longer valid.
For the present case, we can apply the coarsest mesh that still yields a convergent result at a
distance of up to 20 m from the edge.
Let us recall that our simplified models produce finite values for interface forces, even on the
edge; however for purposes of consistency, we will, throughout the following, always be
comparing FE results and M4 analytical solutions only over what has been designated a zone
14
of confidence (i.e. where finite element calculations are not mesh-dependent. Only over such a
zone therefore are our models able to be validated. Nevertheless, the pertinence of finite
values on the edges of M4 interface forces may be improved by complementary approach, e.g.
experimentation (see Caron et al. (1999)).
As a conclusion to this section on numerical aspects, the interest of finite element
generalised internal forces has been clearly demonstrated: the stresses are calculated at
exactly the interface and convergence of the results is better ensured.
In this section, our goal is to compare, over the zone of confidence, finite element results with
M4 analytical solutions through, in particular, a 3D stress fields comparison (throughout the
laminate thickness), and then the interface stresses (finite element generalised forces). Finally,
a study on energy distributions is conducted. It should be pointed out that the M4_5n
analytical solution to the micro-cracked (0,90)s laminate can be found in Carreira (1998).
15
Fig 9a : Shear stress distributions as a function of z/e for different distances from the edge.
Free edge problem
Fig 9b : Shear stress distributions as a function of z/e for different distances from the edge.
Micro-cracked cell
16
6.2 Interface shear and normal stresses
We recall that the finite element shear and normal stresses considered herein stem from the
derivation of the finite element membranar and shear forces (finite element generalised
forces).
The form of the shear stress, as determined analytically by the M4_5n, is well reproduced, for
both of the two boundary conditions, except at the vicinity of the edge (see Figures 10). As
regards M4_2n+1, the shear stress distributions are not as accurate as those provided by the
M4_5n. Nevertheless, they do remain quite satisfactory, particularly with respect to the
maximum values at the edges. The comparisons with Pagano's results also take place below.
We have also plotted the distributions of the 0 ,90 and 90 ,90 normal stresses, as given by the
M4_5n, respectively at the 0/90 and 90/90 interface (see Figures 11). The high level of
correlation between the curves is well established except, perhaps, at the 0/90 interface where
a singularity has been noticed. In the case of micro-cracking, the finite element normal
stresses also indicate singularities at the edge.
For the free edge problem, we have plotted the results given by Pagano's model which exactly
verify the edge conditions; consequently, the shear stress is equal to zero on y=b. The
maximum shear stress value occurs inside the ply and is less than that obtained by finite
elements. Concerning the normal stress, the correlation between M4_5n (no normal stress in
M4_2n+1) and Pagano's results is better, and this is for both interfaces. It is a very important
point that our simplified models agree quite closely (for 2n+1) or even better (for 5n) with
finite elements (in the confidence zone) than a more sophisticated model.
Moreover our models provide a very useful finite value on the edge, value to be improved by
a delamination criterion (see Caron et al. (1999)). These two points reveal how well designed
these simplified models are in studying edge effects.
In Figure 12, we have added results for a (45)s, stemming from 3D calculations (that have
not been described in this paper). The conclusions are similar and justify our choice of more
straightforward 2D studies.
17
Fig 10a : Shear stresses at the 0/90 interface. Free edge problem
18
Fig 11a : Normal stresses at the 0/90 and 90/90 interfaces. Free edge problem
Fig 11b : Normal stresses at the 0/90 and 90/90 interfaces. Micro-cracked cell
19
Fig 12 : Shear stresses at the +45/-45 interface in a (45)s. Free edge problem
i
where S mnop are the components of the ply i compliance tensor.
(w )dy + (w )
b b
W 5n = 5 n 90
+ wM5 n, + w5n + w5n + wM5 n, + w5n + w5 n dy
90 90 90
5n0 0 0 0
M M (17)
y =0 y =0
where appearing in this order, the elastic energy due to membranar stresses, the membranar-
normal coupling energy, the normal stress energy, and the shear stress energy. The expression
of the shear stress energy is given as an example:
20
1 i i e i i ,i +1
( i 1,i
i Q2 S 2323 Q2 + 12 2 2 S 2323 2 2
i i i ,i +1
)
i 1,i
( )
1 e
= 4.
i
w5n (18)
2 1 i e i i ,i +1
+ i Q2
2
( i 1,i i
2 + 2 S 2323 Q2
)
i e i i ,i +1
2
2 + i21,i ( )
5e
At this point, we can define the corresponding finite element energy W EF , where ij
are the finite element stresses values and A el the surface of an element el.
W EF = A
el 90
el
[w EF 90
M + wMEF,
90
+ wEF
90
+ wEF
90
]+ A [w
el 0
el EF 0
M
0 0
+ wMEF, + wEF + wEF
0
] (19)
where for each ply, we have introduced, membranar stresses energy, coupling energy, normal
stress energy and shear stresses energy which is for example: wEF = 2 S 2323
i
i23 .
0
( ) 2
Table 2 compares the various energy contributions for the free edge problem in (0,90)s
laminates. Differences between the two plies have also been studied :
21
And for the microcracked cell,
I.E.M4_5n = 7.84E-02 J
I.E.FE = 7.42E-02 J
The only difference with 5n energy concerns the expression of the shear stress energy :
i 1 ei
( ) (
w2 n +1 = 4 . i ,i +1 i 1,i S i 3 3 i ,i +1 i 1,i +
2 12
) (
e i i ,i +1
) ( )
+ i 1,i S i 3 3 i ,i +1 + i 1,i (21)
4
We should emphasise that the coupling energy is still assumed to be negligible and that the
normal interface energy is not present in this model (as a consequence of not introducing the
corresponding generalised force).
If we compare the M4_2n+1 shear stress energy and the finite element interface
energy for the free edge problem :
w2 n +1 = 7.12E-02 J vs I.E.FE = 6.78E-02 J
and for the microcracked laminate :
w2 n +1 = 7.62E-02 J vs I.E.FE = 7.42E-02 J
we would like to highlight once again the value of this I.E. concept, even for this simple
model.
The energetic analysis we have performed in this section provides a better
understanding of our model descriptions as well as an explanation of the meaning of a
simplified model (e.g. just as the Love-Kirchhoff plate model is a simplified Reissner plate
model, we can consider M4_2n+1 as a simplified M4_5n) : when a generalised force vanishes
in a simplified model, this means that the associated energy is simply transferred or
distributed into the other terms . In this way, a concept of interface energy, as the sum of
energies due to shear and normal stresses, has been defined and validated.
22
7. Conclusion
This paper deals with the validation of simplified models which involve one kinematic
per layer. We developed such a model that we named Multiparticle Models of Multilayered
Materials (M4). These models allow introducing out-of-plane stresses (i.e. shear and normal
stresses) at the interfaces of a laminate. We have examined in depth two of these models : the
M4_5n and the M4_2n+1. The first one can be described as a stacking sequence of Reissner-
Mindlin plates, the second as a membranar superposition.
The validations are lead by means of finite element calculations in a (0,90)s laminate
submitted to tension. Two boundary conditions were considered : free edge and
microcracking. We also compare with results stemming from Pagano's works.
First of all, the validation procedure encountered numerical difficulties, since the finite
element stresses are mean element values and not calculated at the right interface. For this
reason, we introduced what we called finite element generalised forces, which are deduced
from the 3D equilibrium equations and actually represent interface stresses. Next, a
convergence study was conducted using these generalised forces, which are more relevant and
stable tools. We could then focus on the singular behaviour of the stresses when approaching
the edge or the microcrack; the maximum value of stresses thus depends upon the level of
mesh refinement. Considering the nature of laminate edges (pulled out fibres, defects due to
the elaboration process), we have proposed to define a zone of confidence (excluding a region
where the calculated stresses have no meaning in relation to material heterogeneity) over
which the convergence of the finite element generalised forces is ensured.
In comparing M4_5n and finite element results in this zone of confidence, the
following conclusions could be established : 3D stress fields are accurately reproduced even
with critical boundary conditions (i.e. in the vicinity of a micro-crack) and the energy
contributions in each ply, associated to the different stresses are calculated extremely well.
Validation was also carried out with the M4_2n+1 : due to the lower degree of the
polynomials approximating the stress fields, this model is obviously less precise in describing
stress distributions. The conclusion is that our simplified models seem to be very attractive,
because more simple and more convenient for the study of edge effects, than more
sophisticated ones.
With respect to the various energy contributions, our work has led to defining an interface
energy by summing the two energies related to shear and normal stresses, respectively. A
23
comparison with the same 3D energy proves to be very close: this point emphasises the fact
that energetic approaches constitute a promising way to propose delamination criteria even
with a simple model such as M4_2n+1.
Once again, we would like to insist that in the present work, all the conclusions and
comparisons drawn between FE and analytical solutions concern the zone of confidence of
laminates : nonetheless, the pertinence of the finite values on the edges of M4 interface forces
shall be improved by an experimental approach (Caron et al. (1999)).
REFERENCES
Allix O., 1992. Interlaminar interface modelling for the prediction of delamination.
Composite Structures, Vol. 22, p. 235-242.
Caron J.F., Carreira R.P. , Diaz Diaz A., 1999. Critres dinitiation de dlaminage dans les
stratifis. C. R. Acad. Sci. Paris, t. 327, Srie II b, p. 1291.
Caron J.F., Ehrlacher A., 1997. Modelling the kinetics of transverse cracking in composite
laminates. Composites Science and Technology, Vol. 57, p. 1261.
Caron J.F., Ehrlacher A., 1998. Modelling of fatigue microcracking kinetics in crossply and
experimental validation. Composites Science and Technology, Vol. 59, p. 1349.
Carreira R.P., 1998. Validation par lments finis des Modles Multiparticulaires de
Matriaux Multicouches (M4). PhD Thesis, Ecole Nationale des Ponts et Chausses,
France.
Chabot A., 1997. Analyse des efforts linterface entre les couches des matriaux composites
laide de Modles Multiparticulaires de Matriaux Multicouches (M4). PhD Thesis,
Ecole Nationale des Ponts et Chausses, France.
Garett K. W., Bailey J. E., 1977. Multiple transverse fracture in 90 cross-ply laminates of a
glass fibre-reinforced polyester. Journal of Materials Science, Vol. 12, p. 157.
Kassapoglou C., Lagace P. A., 1986. An efficient method for the calculation of interlaminar
stresses in composite materials. Journal of Applied Mechanics, Vol. 53, p. 745.
Lcuyer F., 1991. Etude des effets de bord dans les structures minces multicouches. PhD
Thesis, Universit Paris VI, France.
Lee J. W., Daniel I. M., 1990. Progressive transverse cracking of crossply composite
laminates. Journal of Composite Materials, Vol. 24, p. 1225.
Leguillon D., 1998. A method based on singularity theory to predict edge delamination of
laminates. International Journal of Fracture, Vol. 100, p. 105.
24
Naciri T., Ehrlacher A., Chabot A., 1998. Interlaminar stress analysis with a new
Multiparticle Modelization of Multilayered Materials (M4). Composites Science and
Technology, Vol 58, p. 337.
Pagano N. J., 1978. Stress fields in composite laminates. International Journal of Solids and
Structures, Vol. 14, p. 385-400.
Raju I. S., Crews J. H., 1981. Interlaminar stress singularities at a straight free edge in
composite laminates. Computers and Structures, Vol. 14, No 1-2, p. 21.
Reissner E., 1950. On a variational theorem in elasticity. Journal of Mathematics and Physics,
Vol. 29, p. 90.
Robbins D. H., Reddy J. N., 1993. Modelling of thick composites using a layerwise laminate
theory. International Journal for Numerical Methods in Engineering, Vol. 36, p. 655.
Shah C. G., Krishna Murty A. V., 1991. Analysis of edge delaminations in laminates through
combined use of quasi-three dimensional eight-noded, two-noded and transition elements.
Computers and Structures, Vol. 39, No 3-4, p. 231.
Steif P. S., 1983 in Ogin et al., 1985 : Matrix cracking and stiffness reduction during the
fatigue of a (0,90)s GFRP laminate. Composites Science and Technology, Vol 22, p. 23.
Wang A. S. D., Crossman F. W., 1977. Some new results on edge effect in symmetric
composite laminates. Journal of Composite Materials, Vol. 11, p. 92.
Wang S. S., Choi I., 1982a. Boundary-layer effects in composite laminates. Part 1 : Free edge
stress singularities. Journal of Applied Mechanics, Vol. 49, p. 541.
Wang S. S., Choi I., 1982b. Boundary-layer effects in composite laminates. Part 2 : Free edge
stress solutions and basic characteristics. Journal of Applied Mechanics, Vol. 49, p. 549.
25
26